Hasil untuk "Pathology"

Fei Tan, Ashok Vardhan Addala, Bruno Astuto Arouche Nunes et al.

Medical imaging datasets often suffer from class imbalance and limited availability of pathology-rich cases, which constrains the performance of machine learning models for segmentation, classification, and vision-language tasks. To address this challenge, we propose POWDR, a pathology-preserving outpainting framework for 3D MRI based on a conditioned wavelet diffusion model. Unlike conventional augmentation or unconditional synthesis, POWDR retains real pathological regions while generating anatomically plausible surrounding tissue, enabling diversity without fabricating lesions. Our approach leverages wavelet-domain conditioning to enhance high-frequency detail and mitigate blurring common in latent diffusion models. We introduce a random connected mask training strategy to overcome conditioning-induced collapse and improve diversity outside the lesion. POWDR is evaluated on brain MRI using BraTS datasets and extended to knee MRI to demonstrate tissue-agnostic applicability. Quantitative metrics (FID, SSIM, LPIPS) confirm image realism, while diversity analysis shows significant improvement with random-mask training (cosine similarity reduced from 0.9947 to 0.9580; KL divergence increased from 0.00026 to 0.01494). Clinically relevant assessments reveal gains in tumor segmentation performance using nnU-Net, with Dice scores improving from 0.6992 to 0.7137 when adding 50 synthetic cases. Tissue volume analysis indicates no significant differences for CSF and GM compared to real images. These findings highlight POWDR as a practical solution for addressing data scarcity and class imbalance in medical imaging. The method is extensible to multiple anatomies and offers a controllable framework for generating diverse, pathology-preserving synthetic data to support robust model development.

Eric Zimmermann, Julian Viret, Michal Zelechowski et al.

In recent years, a standard computational pathology workflow has emerged where whole slide images are cropped into tiles, these tiles are processed using a foundation model, and task-specific models are built using the resulting representations. At least 15 different foundation models have been proposed, and the vast majority are trained exclusively with tiles using the 20$\times$ magnification. However, it is well known that certain histologic features can only be discerned with larger context windows and requires a pathologist to zoom in and out when analyzing a whole slide image. Furthermore, creating 224$\times$224 pixel crops at 20$\times$ leads to a large number of tiles per slide, which can be gigapixel in size. To more accurately capture multi-resolution features and investigate the possibility of reducing the number of representations per slide, we propose a region-level mixing encoder. Our approach jointly fuses image tile representations of a mixed magnification foundation model using a masked embedding modeling pretraining step. We explore a design space for pretraining the proposed mixed-magnification region aggregators and evaluate our models on transfer to biomarker prediction tasks representing various cancer types. Results demonstrate cancer dependent improvements in predictive performance, highlighting the importance of spatial context and understanding.

Kang Yu, Dingyu Wang, Zimu Yuan et al.

Musculoskeletal disorders represent a leading cause of global disability, creating an urgent demand for precise interpretation of medical imaging. Current artificial intelligence (AI) approaches in orthopedics predominantly rely on task-specific, supervised learning paradigms. These methods are inherently fragmented, require extensive annotated datasets, and often lack generalizability across different modalities and clinical scenarios. The development of foundation models in this field has been constrained by the scarcity of large-scale, curated, and open-source musculoskeletal datasets. To address these challenges, we introduce OrthoFoundation, a multimodal vision foundation model optimized for musculoskeletal pathology. We constructed a pre-training dataset of 1.2 million unlabeled knee X-ray and MRI images from internal and public databases. Utilizing a Dinov3 backbone, the model was trained via self-supervised contrastive learning to capture robust radiological representations. OrthoFoundation achieves state-of-the-art (SOTA) performance across 14 downstream tasks. It attained superior accuracy in X-ray osteoarthritis diagnosis and ranked first in MRI structural injury detection. The model demonstrated remarkable label efficiency, matching supervised baselines using only 50% of labeled data. Furthermore, despite being pre-trained on knee images, OrthoFoundation exhibited exceptional cross-anatomy generalization to the hip, shoulder, and ankle. OrthoFoundation represents a significant advancement toward general-purpose AI for musculoskeletal imaging. By learning fundamental, joint-agnostic radiological semantics from large-scale multimodal data, it overcomes the limitations of conventional models, which provides a robust framework for reducing annotation burdens and enhancing diagnostic accuracy in clinical practice.

Chunze Yang, Wenjie Zhao, Yue Tang et al.

Precision pathology relies on detecting fine-grained morphological abnormalities within specific Regions of Interest (ROIs), as these local, texture-rich cues - rather than global slide contexts - drive expert diagnostic reasoning. While Vision-Language (V-L) models promise data efficiency by leveraging semantic priors, adapting them faces a critical Granularity Mismatch, where generic representations fail to resolve such subtle defects. Current adaptation methods often treat modalities as independent streams, failing to ground semantic prompts in ROI-specific visual contexts. To bridge this gap, we propose the Hierarchical Adaptation and Alignment Framework (HAAF). At its core is a novel Cross-Level Scaled Alignment (CLSA) mechanism that enforces a sequential calibration order: visual features first inject context into text prompts to generate content-adaptive descriptors, which then spatially guide the visual encoder to spotlight anomalies. Additionally, a dual-branch inference strategy integrates semantic scores with geometric prototypes to ensure stability in few-shot settings. Experiments on four benchmarks show HAAF significantly outperforms state-of-the-art methods and effectively scales with domain-specific backbones (e.g., CONCH) in low-resource scenarios.

Tiantian Sheng, Hang Su, Lu Yao et al.

Abstract Docetaxel is a widely used first-line treatment for castration-resistant prostate cancer (CRPC). RhoB, a member of the Rho GTPase family, plays a major role in prostate cancer metastasis by modulating the PI3K-AKT signaling pathway. It is crucial in regulating cytoskeletal reassembly, cell migration, focal adhesion (FA) dynamics. To investigate RhoB’s function in prostate cancer, CRISPR/Cas9 gene editing technique was utilized to knock out the RhoB gene in prostate cancer cells. Successful gene editing was confirmed by using T7 endonuclease I (T7EI) assays and Sanger sequencing. Knocking out RhoB enhanced epithelial–mesenchymal transition (EMT) and decreased the IC50 value of docetaxel in RhoB-knockout PC-3 cells. This suggests increased sensitivity to docetaxel. Furthermore, RhoB knockout prompted the migration and invasion of prostate cancer cells, effects that were reversed upon RhoB overexpression. Interestingly, RhoB status did not significantly influence the cell cycle of prostate cancer cells. RNA sequencing of PC-3 cells with either overexpressed or knock-out RhoB revealed that RhoB regulates pathways involved in FA, ECM receptor interaction, and PI3K-AKT signaling. These pathways directly influence the EMT process, cell migration, and invasion in prostate cancer cells. Notably, RhoB overexpression activated PI3K-AKT signaling when PC-3 cells were treated with low concentration of DTXL (50 nM, 72 h). This activation reduced DTXL’s cytotoxicity, suggesting may confer chemoresistance via PI3K-AKT pathway activation.

Guoguang Lv, Shiyu Zhang, Ting Zhang et al.

A 71 year-old male was diagnosed of epidermoid cyst located in diploe and cerebellum. The skull part was found firstly and kept steady for more than 5 years. The cerebellar part was found nearby when dizziness and vomit happened. The patient has gone through a traumatic brain injury 4 decades ago. All lesions were resected totally. Interestingly the dura mater was confirmed intact without any leakage into subdural space. Imaging and pathological materials are collected completely. Conclusion: We report a case that EC was found in both skull and cerebellum whereas the dural mater was intact. Epidermoid cell migration or infiltration are possible explanations. Gross total resection is advanced for better clinical outcome.

Yicheng Song, Tiancheng Lin, Die Peng et al.

Various multi-instance learning (MIL) based approaches have been developed and successfully applied to whole-slide pathological images (WSI). Existing MIL methods emphasize the importance of feature aggregators, but largely neglect the instance-level representation learning. They assume that the availability of a pre-trained feature extractor can be directly utilized or fine-tuned, which is not always the case. This paper proposes to pre-train feature extractor for MIL via a weakly-supervised scheme, i.e., propagating the weak bag-level labels to the corresponding instances for supervised learning. To learn effective features for MIL, we further delve into several key components, including strong data augmentation, a non-linear prediction head and the robust loss function. We conduct experiments on common large-scale WSI datasets and find it achieves better performance than other pre-training schemes (e.g., ImageNet pre-training and self-supervised learning) in different downstream tasks. We further show the compatibility and scalability of the proposed scheme by deploying it in fine-tuning the pathological-specific models and pre-training on merged multiple datasets. To our knowledge, this is the first work focusing on the representation learning for MIL.

Boqi Chen, Cédric Vincent-Cuaz, Lydia A. Schoenpflug et al.

Vision foundation models (FMs) are accelerating the development of digital pathology algorithms and transforming biomedical research. These models learn, in a self-supervised manner, to represent histological features in highly heterogeneous tiles extracted from whole-slide images (WSIs) of real-world patient samples. The performance of these FMs is significantly influenced by the size, diversity, and balance of the pre-training data. However, data selection has been primarily guided by expert knowledge at the WSI level, focusing on factors such as disease classification and tissue types, while largely overlooking the granular details available at the tile level. In this paper, we investigate the potential of unsupervised automatic data curation at the tile-level, taking into account 350 million tiles. Specifically, we apply hierarchical clustering trees to pre-extracted tile embeddings, allowing us to sample balanced datasets uniformly across the embedding space of the pretrained FM. We further identify these datasets are subject to a trade-off between size and balance, potentially compromising the quality of representations learned by FMs, and propose tailored batch sampling strategies to mitigate this effect. We demonstrate the effectiveness of our method through improved performance on a diverse range of clinically relevant downstream tasks.

Ali Mammadov, Loïc Le Folgoc, Guillaume Hocquet et al.

Digital pathology has revolutionized the field by enabling the digitization of tissue samples into whole slide images (WSIs). However, the high resolution and large size of WSIs present significant challenges when it comes to applying Deep Learning models. As a solution, WSIs are often divided into smaller patches with a global label (\textit{i.e., diagnostic}) per slide, instead of a (too) costly pixel-wise annotation. By treating each slide as a bag of patches, Multiple Instance Learning (MIL) methods have emerged as a suitable solution for WSI classification. A major drawback of MIL methods is their high variability in performance across different runs, which can reach up to 10-15 AUC points on the test set, making it difficult to compare different MIL methods reliably. This variability mainly comes from three factors: i) weight initialization, ii) batch (shuffling) ordering, iii) and learning rate. To address that, we introduce a Multi-Fidelity, Model Fusion strategy for MIL methods. We first train multiple models for a few epochs and average the most stable and promising ones based on validation scores. This approach can be applied to any existing MIL model to reduce performance variability. It also simplifies hyperparameter tuning and improves reproducibility while maintaining computational efficiency. We extensively validate our approach on WSI classification tasks using 2 different datasets, 3 initialization strategies and 5 MIL methods, for a total of more than 2000 experiments.

Quentin Posva

We construct pathological examples of MMP singularities in every positive characteristic using quotients by $α_p$-actions. In particular, we obtain non-$S_3$ terminal singularities, as well as locally stable (respectively stable) families whose general fibers are smooth (respectively klt, Cohen--Macaulay and $F$-injective) and whose special fibers are non-$S_2$. The dimensions of these examples are bounded below by a linear function of the characteristic.

A. Ghenim, D. Brahim, I. Yaich et al.

Introduction Internet use can become uncontrollable, leading to physical and psychological suffering and what is known as cyberaddiction. Objectives To assess the frequency of cyberaddiction in a population of young doctors. Methods We conducted a cross-sectional, descriptive study of a population of young doctors. We collected socio-professional and medical data using a Google Forms self-questionnaire. The Young scale was recommended for screening for cyberaddiction. A score ≥5 indicates Internet addiction. The Hospital Anxiety and Depression Scale (HAD) was adopted to reveal anxiety-depressive disorders. Results A total of 45 physicians responded to our survey. The mean age was 29.93±4.8 years. The sex ratio (M/F) was 0.3. Participants were single in 69% of cases. Residents represented 64% of the population. Physicians were family medicine residents in 11% of cases. The mean Young’s score was 3.13±1.97/8. Cyberaddiction was noted in 24% of cases. A definite anxiety-depressive disorder was found in 6.7% and 13.3% of cases respectively. Internet addiction was significantly associated with female gender (p<0.05) and a positive HAD (A) score (p=0.03). Conclusions According to the results of our study, cyberaddiction is common among medical staff. A preventive strategy is needed to counter the harmful effects of this addiction. Disclosure of Interest None Declared

Shimaa A. Abdelbaky, Zakaria M. Zaky, Doha Yahia et al.

Contamination of the environment with nano- and microplastic particles exerts a threatening impact on the aquatic ecosystems and sustainable catfish aquaculture. The presence of nanoplastics has been found to have a detrimental impact on both aquatic and terrestrial ecosystems. The present study examines the effect of polystyrene nanoplastics (PS NPs) on the DNA, erythrocytes, oxidative status and renal histology of catfish, in addition to the potential protective effects of <i>Chlorella vulgaris</i> bioremediation and selenium to hinder this effect. Six equal groups of fish were used as follows: Group 1 served as a control group and received water free from PS NPs; Group 2 was exposed to PS NPs at a concentration of 5 mg/L; Group 3 was exposed to PS NPs (5 mg/L) + selenium (1 mg/kg diet); Group 4 was exposed to PS NPs (5 mg/L) + <i>C. vulgaris</i> (25 g/kg diet); Group 5 was supplemented with <i>C. vulgaris</i> (25 g/kg diet); and Group 6 was supplemented with selenium (1 mg/kg diet). The exposure period was 30 days. The results indicated that PS NPs induced oxidative stress by significantly elevating malondialdehyde activities and slightly reducing antioxidant biomarkers, resulting in DNA damage, increased frequency of micronuclei, erythrocyte alterations, and numerous histopathological alterations in kidney tissue. Selenium and <i>C. vulgaris</i> significantly ameliorated the oxidative/antioxidant status, reducing DNA damage, micronucleus frequency, erythrocyte alterations, and improving the morphology of kidney tissue. Nevertheless, further research is needed for a profound understanding of the mechanism behind the toxicity of nano-microplatics in aquatic systems.

Omar S. M. El Nahhas, Georg Wölflein, Marta Ligero et al.

Deep Learning (DL) can predict biomarkers directly from digitized cancer histology in a weakly-supervised setting. Recently, the prediction of continuous biomarkers through regression-based DL has seen an increasing interest. Nonetheless, clinical decision making often requires a categorical outcome. Consequently, we developed a weakly-supervised joint multi-task Transformer architecture which has been trained and evaluated on four public patient cohorts for the prediction of two key predictive biomarkers, microsatellite instability (MSI) and homologous recombination deficiency (HRD), trained with auxiliary regression tasks related to the tumor microenvironment. Moreover, we perform a comprehensive benchmark of 16 approaches of task balancing for weakly-supervised joint multi-task learning in computational pathology. Using our novel approach, we improve over the state-of-the-art area under the receiver operating characteristic by +7.7% and +4.1%, as well as yielding better clustering of latent embeddings by +8% and +5% for the prediction of MSI and HRD in external cohorts, respectively.

Gregory W. Charville

Karen Tew, Jesse Kresak, Stacy G. Beal

Kenji Tsuji, Hiroyuki Nakanoh, Kensaku Takahashi et al.

Rationale &amp; Objective: Assessment of kidney biopsies provides crucial information for diagnosis and disease activity, as well as prognostic value. Kidney-biopsy specimens occasionally contain veno-muscular complex (VMC), which consists of muscle tissues around the kidney venous system in the corticomedullary region. However, the role of VMC and the clinical significance of VMC variants are poorly understood. In the present study, we investigated kidney prognostic values of VMC variants. Study Design: Retrospective cohort study. Setting &amp; Participants: Among 808 patients who underwent a kidney biopsy from 2011 to 2019, 246 patients whose kidney biopsy specimens contained VMC were enrolled. Predictors: VMC variants; inflammatory-VMC (an infiltration of ≥80 inflammatory cells/mm2-VMC area) and VMC hypertrophy (hyper-VMC, a VMC average width ≥850 μm), and the interstitial fibrosis/tubular atrophy (IFTA) score. Outcomes: A decline in estimated glomerular filtration rate (eGFR) ≥40% from the baseline or commencement of kidney replacement therapy. Analytical Approach: Cox proportional hazards model. Results: Among 246 patients with data on VMC, mean baseline eGFR was 56.0±25.6 ml/min per 1.73 m2; 80 had high inflammatory-VMC, and 62 had VMC hypertrophy. There were 51 kidney events over median follow-up of 2.5 years. We analyzed 2 VMC variants. Multivariable logistic regression analysis revealed that eGFR negatively correlated with the presence of both inflammatory-VMC and hyper-VMC. A Cox proportional hazards analysis revealed that inflammatory-VMC (but not hyper-VMC) was independently associated with the primary outcome after adjustments for known risk factors of progression, including proteinuria, eGFR, and the interstitial fibrosis/tubular atrophy (IFTA) score (hazard ratio, 1.97; 95% confidence interval, 1.00-3.91). Limitations: Single-center study and small sample size. Conclusions: Assessment of inflammatory-VMC provides additional kidney prognostic information to known indicators of kidney disease progression in patients who undergo kidney biopsy. Plain-Language Summary: Assessment of kidney biopsies provides crucial information for diagnosis, disease activity, and prognostic value. Kidney-biopsy specimens occasionally contain veno-muscular complex (VMC), which consists of muscle tissues around the kidney venous system. Currently, the role of VMC in kidney health and diseases and the clinical significance of VMC variants are poorly understood. In the present study, we have shown that an infiltration of ≥80 inflammatory cells/mm2-VMC area (inflammatory-VMC) is independently associated with kidney disease progression after adjustments for known risk factors of progression. Therefore, assessment of inflammatory-VMC provides additional kidney prognostic information to known indicators of kidney disease progression in patients who undergo kidney biopsy.

Ana Susac, Maja Planinic, Andreja Bubic et al.

Eye tracking can provide valuable insights into how students use different representations to solve problems and can be a useful tool for measuring the integration of information from multiple representations. In this study, we measured the eye movements of 60 university students while solving two PISA items that contain graphs taken from mathematics and science assessments with the aim of studying the difference in visual attention between students who correctly and incorrectly identify graphs from a verbal description. We were particularly interested in the differences in the integration of information from different representations (text, graphs, and picture) between students who were successful or unsuccessful in solving items. The results suggest that students who solved the items correctly tend to solve the items longer than their counterparts who did not solve the items correctly. Analysis of eye tracking data suggests that students who solved science item correctly analyzed the graph for significantly longer time and had significantly longer average fixation time. This finding suggests that a careful analysis of graphs is crucial for the correct solution of PISA items used in this study. Furthermore, the results showed that students who solved the mathematics item correctly had significantly higher number of transitions between graphs and picture, which indicates a greater integration of information from two different representations. This indicates that these types of items require a lot of time and effort to complete, probably because solving them requires a lot of steps, which is cognitively demanding. We also found that the average fixation durations for different representations may vary for different items, indicating that it is not always equally difficult to extract necessary information from different types of representations. The results of this study suggest that instructors may be able to improve their teaching methods by considering the importance of individual representations (e.g., texts, graphs, and pictures) and the integration of information from multiple sources.

Oleksii Nikolaienko, Hans P. Eikesdal, Elisabet Ognedal et al.

Abstract Background Normal cell BRCA1 epimutations have been associated with increased risk of triple-negative breast cancer (TNBC). However, the fraction of TNBCs that may have BRCA1 epimutations as their underlying cause is unknown. Neither are the time of occurrence and the potential inheritance patterns of BRCA1 epimutations established. Methods To address these questions, we analyzed BRCA1 methylation status in breast cancer tissue and matched white blood cells (WBC) from 408 patients with 411 primary breast cancers, including 66 TNBCs, applying a highly sensitive sequencing assay, allowing allele-resolved methylation assessment. Furthermore, to assess the time of origin and the characteristics of normal cell BRCA1 methylation, we analyzed umbilical cord blood of 1260 newborn girls and 200 newborn boys. Finally, we assessed BRCA1 methylation status among 575 mothers and 531 fathers of girls with (n = 102) and without (n = 473) BRCA1 methylation. Results We found concordant tumor and mosaic WBC BRCA1 epimutations in 10 out of 66 patients with TNBC and in four out of six patients with estrogen receptor (ER)-low expression (< 10%) tumors (combined: 14 out of 72; 19.4%; 95% CI 11.1–30.5). In contrast, we found concordant WBC and tumor methylation in only three out of 220 patients with 221 ER ≥ 10% tumors and zero out of 114 patients with 116 HER2-positive tumors. Intraindividually, BRCA1 epimutations affected the same allele in normal and tumor cells. Assessing BRCA1 methylation in umbilical WBCs from girls, we found mosaic, predominantly monoallelic BRCA1 epimutations, with qualitative features similar to those in adults, in 113/1260 (9.0%) of individuals, but no correlation to BRCA1 methylation status either in mothers or fathers. A significantly lower fraction of newborn boys carried BRCA1 methylation (9/200; 4.5%) as compared to girls (p = 0.038). Similarly, WBC BRCA1 methylation was found less common among fathers (16/531; 3.0%), as compared to mothers (46/575; 8.0%; p = 0.0003). Conclusions Our findings suggest prenatal BRCA1 epimutations might be the underlying cause of around 20% of TNBC and low-ER expression breast cancers. Such constitutional mosaic BRCA1 methylation likely arise through gender-related mechanisms in utero, independent of Mendelian inheritance.

Rebecca C. Larson, Michael C. Kann, Charlotte Graham et al.

Abstract Chimeric Antigen Receptor (CAR) T cells directed to B cell maturation antigen (BCMA) mediate profound responses in patients with multiple myeloma, but most patients do not achieve long-term complete remissions. In addition, recent evidence suggests that high-affinity binding to BCMA can result in on-target, off-tumor activity in the basal ganglia and can lead to fatal Parkinsonian-like disease. Here we develop CAR T cells against multiple myeloma using a binder to targeting transmembrane activator and CAML interactor (TACI) in mono and dual-specific formats with anti-BCMA. These CARs have robust, antigen-specific activity in vitro and in vivo. We also show that TACI RNA expression is limited in the basal ganglia, which may circumvent some of the toxicities recently reported with BCMA CARs. Thus, single-targeting TACI CARs may have a safer toxicity profile, whereas dual-specific BCMA-TACI CAR T cells have potential to avoid the antigen escape that can occur with single-antigen targeting.

Daniela P. Schacherer, Markus D. Herrmann, David A. Clunie et al.

Background and Objectives: Reproducibility is a major challenge in developing machine learning (ML)-based solutions in computational pathology (CompPath). The NCI Imaging Data Commons (IDC) provides >120 cancer image collections according to the FAIR principles and is designed to be used with cloud ML services. Here, we explore its potential to facilitate reproducibility in CompPath research. Methods: Using the IDC, we implemented two experiments in which a representative ML-based method for classifying lung tumor tissue was trained and/or evaluated on different datasets. To assess reproducibility, the experiments were run multiple times with separate but identically configured instances of common ML services. Results: The AUC values of different runs of the same experiment were generally consistent. However, we observed small variations in AUC values of up to 0.045, indicating a practical limit to reproducibility. Conclusions: We conclude that the IDC facilitates approaching the reproducibility limit of CompPath research (i) by enabling researchers to reuse exactly the same datasets and (ii) by integrating with cloud ML services so that experiments can be run in identically configured computing environments.